Many modern artificial intelligence (AI) systems, including both real physical robots and animat-based models are structured using intelligent agents. A key challenge for building large and complex AI systems is to manage the agent interactions in an appropriate architecture that supports complexity in a scalable and hierarchical manner. We review various agent architectures for both physical and simulated robot systems, and show how appropriate agent communications protocols can be developed to support artificially intelligent systems based on communities of interacting agents. As well as reviewing some of the architectures and supporting software tools and technologies, we present our own ideas for a software architecture for managing intelligent agents. We emphasise the importance of being able to incrementally augment the set of agents as new ideas are developed. We describe how key activities such as agent navigation in physical and simulated spaces; agent communication; world state management and sensory integration all need to be managed in an appropriate framework to support individual agents that will take responsibility for tasks and goals.
Keywords: AI; agents; software; architectures; path algorithms.
Full Document Text: PDF version.
Citation Information:
BiBTeX reference:
@techreport{CSTN-054,
title="Software Integration Architectures for Agents",
author="K.A. Hawick and A.P. Gerdelan",
year="2008",
month="May",
series="CSTN-054",
institution="Information and Mathematical Sciences, Massey University",
address="Albany, North Shore 102-904, Auckland, New Zealand"
}
Plain bibitem entry:
\bibitem{CSTN-054}
K.A.Hawick and A.P.Gerdelan, Software Integration Architectures for Agents,
Computational Science Technical Note CSTN-054, Information and Mathematical Sciences,
Massey University, Albany, North Shore 102-904, Auckland, New Zealand, May 2008.